Cutter 8
$ 34 /month
- vCPU
- 4 × dedicated
- RAM
- 8 GB
- Storage
- 160 GB NVMe SSD
- Transfer
- 10 TB
- IPv4 / IPv6
- 1 / /64 routed
Solution
VPS for monitoring: Prometheus, Grafana and uptime checks
Cutter 8 · $34/moCutter 16 · $62/mo
Short answer
A monitoring stack is memory and disk bound. Prometheus with 15-second scrapes across 50 targets and 90 days of retention needs about 8 GB of RAM and 200 GB of NVMe. Put it somewhere other than the infrastructure it watches — monitoring that shares a failure domain with production tells you nothing at the moment you need it most.
What you need
| Resource | What you actually need |
|---|---|
| CPU | 2–4 cores |
| RAM | 8 GB for ~1M active series |
| Disk | NVMe, 160–320 GB for 90-day retention |
| Placement | A different provider or region from what it monitors |
Recommended plans
Cutter 16
$ 62 /month
- vCPU
- 6 × dedicated
- RAM
- 16 GB
- Storage
- 320 GB NVMe SSD
- Transfer
- 15 TB
- IPv4 / IPv6
- 1 / /64 routed
Recommended locations
Location is usually the decision that matters most for this workload — either because latency dominates, or because jurisdiction does.
Why OnionVPS for this
- A cheap instance in a separate jurisdiction gives genuine failure-domain isolation.
- NVMe keeps Prometheus compaction from becoming the loudest thing on the box.
- Instances across 40+ countries let you run blackbox probes from where your users actually are.
- Flat pricing, so retention is a storage decision rather than a per-metric bill.
How to set it up
-
Deploy a Cutter 8 away from your production region
Different country, ideally different provider.
-
Install Prometheus, Alertmanager and Grafana
Docker Compose is the fastest reliable path.
-
Add blackbox probes from several regions
A Skiff 1 in each region running blackbox_exporter is enough.
-
Route alerts somewhere you will see them
Alertmanager to a chat channel, plus a second path for a total outage.
-
Test the alert path deliberately
An untested alerting pipeline is decoration.
Frequently asked questions
How much disk does Prometheus need?
Roughly 1–2 bytes per sample after compression. One million active series at a 15-second scrape interval for 90 days is about 200 GB.
Should monitoring run on the same provider as production?
No. If it shares a failure domain it will go down at exactly the moment you need it. Separate provider or at minimum separate region and jurisdiction.
Can I monitor from multiple countries?
Yes. A $4 Skiff instance per region running blackbox_exporter gives you real geographic probe coverage cheaply.